Fuzzy prediction of the algal blooms in the Orbetello lagoon

The Orbetello lagoon is a shallow brackish waterbody subject to intense and diverse eutrophication (phytoplankton, macroalgae and macrophytes). Periodically a large amount of algae must be artificially removed, their collection and disposal representing a considerable management cost. This paper describes the design of a bloom predictor based on the daily fluctuations of simple water quality parameters such as dissolved oxygen, oxidation–reduction potential, pH and temperature. The task of the fuzzy predictor is to recognise the possibility that a bloom of the macroalgae population is about to occur, based on the changing daily pattern of these variables. The fuzzy predictor is based on a set of fuzzy rules derived from experimental observations and expert knowledge. A 15 month period of hourly data was analysed and used to form the initial knowledge-base. The tests show that the inferential engine has good predictive capabilities, which could be improved when more data become available. The algorithm outlined in the paper has enough generality to be applied to any environmental classification problem where short-term time patterns and longterm statistical variability should be considered jointly.